Should have done this in my previous changes. These options aren't meant
to be exposed to dependee targets. This also allows us to disable/remove
these options on a given target, without depender targets adding them
back.

Bug: b/145758253
Change-Id: Ieaba6824554d74e65d313409916bdf9ce4fd6182
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43928
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>

marl::blocking_call() calls the passed function on another thread.
Attempting to unlock a mutex on a different thread that locked it is
illegal.

Also lock for destruction.

Bug: b/140421726
Change-Id: I87226176bc3a1a454d9c79a139f4d42e51f66814
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43908Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>

Change-Id: I47b09ed307bb2cf4159b375171de0d393a017d3d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43994Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

"ProcessState contains information about an exited process, available after a call to Wait or Run"

As ProcessState is only assigned after `Wait()` has completed, and this is run on another goroutine, it may be nil.
Check before using it.

Change-Id: I319f0703b3547186b53caca8d9d70bc5cd738d34
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43993Reviewed-by: Alexis Hétu <sugoi@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

These used a static for caching the computation, but didn't use a
static initializer. This meant there was a race between the check
for zero and the assignment.

Bug: b/153803432
Change-Id: Id5b91050ced0b8d4811eb32ffed24885816ade7a
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43909Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>

This will also be used by the pipelines for dynamic recompilation.

Bug: b/154158882
Change-Id: I0fe16dbc3bc4eed8ffdd90afcd62f255ca8e4505
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43848
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Fix warning raised by the compiler being too smart for its own good:

  error: loop will run at most once (loop increment never executed) [-Werror,-Wunreachable-code-loop-increment]

Bug: b/153194656
Change-Id: Idca543c13213c6e0bd2d1ff2fdbaa48348f53bf2
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43989Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

LRUCache previously had a complexity of O(n).
Reimplement using a `std::unordered_set` and a linked list to get this reduced to O(1).

Renamed `LRUCache::query()` to `LRUCache::get()`, as this is a more common verb for a cache, and the `query()` suggests it is side effect free (when it actually makes the entry MRU).

Move `LRUCache.hpp` from `src/Device` to `src/System` so it can be tested by `system-unittests`.

Move the logic of `LRUSnapshotCache` into `VkDevice::SamplingRoutineCache`, as this was the only place it was used, and made it exceptionally hard to separate mutex-locked data from non-locked data.

This is part of the work to get our code statically thread-safe-verifiable.

Bug: b/153194656
Change-Id: Ie02888ae6c7ed4066df77d692dfae28c3bc1664d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43489Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Presubmit-Ready: Ben Clayton <bclayton@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Bug: b/145758253
Change-Id: Ifc97d3f833fcf41098ed24aad9cac26f51c28df3
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43870Tested-by: Antonio Maiorano <amaiorano@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

Moved out code from the root CMakeLists into new CMakeLists for each
test and benchmark directory. For Reactor, both the unit test and
benchmark source files are in src/Reactor, so for now, I created folders
for these under tests with a CMakeLists that includes the source files
from src/Reactor. This will be cleaned up eventually when we move
Reactor out of src (b/154013190).

Bug: b/145758253
Change-Id: Ic6bbc394e49d3bdc3105509a19458722fcfe68ee
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43868Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>

They make it unnecessarily easy to write code that violates the mutex lock requirements.

Replace with the `getOrCreate` pattern already used in vk::Device, which keeps the locking internal.

This change fixes a couple of places that we were accesssing data without correct locking (real bugs).

Bug: b/153194656
Change-Id: I132c450594f4042160b575197789bca7f1a5e25f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43650
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

The rules were missing the `Debug/` prefix.

Bug: b/145758253
Change-Id: Ia5b9e4f162c6f685e15bf26ab492dcebc93b844d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43850Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>

According to the Vulkan 1.2 spec:
"For fixed-point depth buffers, fragment depth values are
 always limited to the range [0,1] by clamping after depth
 bias addition is performed. Fragment depth values are
 clamped even when the depth buffer uses a floating-point
 representation."

So, unless the VK_EXT_depth_range_unrestricted extension
is enabled, clamping should always occur.

Bug: b/152634772
Change-Id: I2aee2aeb91efed7399580689fc0ffc0634b2f487
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43869
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>

Add CMakeLists.txt to the following folders under src that create the
following targets:

Device -> vk_device
Pipeline -> vk_pipeline
WSI -> vk_wsi
System -> vk_system
Vulkan -> vk_swiftshader

Dependencies (by folder):

System --> n/a
Pipeline --> System Device* Vulkan*
WSI --> System Pipeline Vulkan*
Device --> System Pipeline Vulkan*
Vulkan --> Device Pipeline WSI System

* = dependency by include directory only (should fix this in the future)

Also make system-unittests and system-benchmarks depend on vk_system
rather than build part of System source files.

Bug: b/145758253
Change-Id: I9cce59ca90d91601696f6195326b5fd7dde517d5
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43690
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

`RoutineCache` was not used, remove it.
Rename `RoutineCacheT` to the now unused name `RoutineCache`.

Bug: b/126126820
Change-Id: I656c1727446754b2063033806634678b14485b6f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43816Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Prefer composition over inheritance.
Makes it clear that the virtual destructors on these processors are unnecessary.

Bug: b/126126820
Change-Id: I062c06191ec1cdfb81f2a2773a976bf5c596e658
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43817Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>

This implements external memory using a Zircon VMO handle
that can be transferred between Fuchsia processes easily.

Bug: b/140419396
Change-Id: I81cecec35b218eb22f3318ddec7b790b89e5ce09
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/36168Tested-by: David Turner <digit@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

The Modf and Frexp instructions from the GLSL.std.450 extended SPIR-V
instruction set take a pointer argument to write one of their results
to. This makes them the only arithmetic instructions which need to know
how to explicitly access memory.

This change replaces the partial duplication of store logic with a call
to the underlying implementation of OpStore. To support storing
intermediate values not associated with SPIR-V objects, the Operand
class can now also wrap an independent Intermediate instance.

Bug: b/153641251
Change-Id: Iebab43640b45ed6c27a77576168481d1a27158b6
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43728
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

The new Store() helper function can store Operand instances independent
from SPIR-V instructions. This allows reuse of this logic for other
instructions that need to store to memory, like Modf and Frexp.

Bug: b/153641251
Change-Id: I453bb7cd24ba26b9a23d73568dc3374a52a36073
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43695
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>

The Operand class is a low-level abstraction of rvalues and constants.
It should not carry the SPIR-V type ID. We only need the size in
components.

Bug: b/129000021
Change-Id: I6cb3ed6341b1ccf5ef759075d7410ba447617c8b
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43693
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Bug: b/129000021
Change-Id: I36401de649eb53474ca74acb069351ce37f7529f
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43828
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

Avoid directly accessing SPIR-V instruction words. The helper methods
provide self-explanatory semantics so we can eliminate local variables
to store these IDs.

Bug: b/129000021
Change-Id: Ie42782d53b9c24014b6a1b1f51b82085b6c2ebef
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43694
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>

The defining instruction already contains the type ID.

Bug: b/129000021
Change-Id: I0b29fda73964f5f4a73181a61a0f30cd1c47f404
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43692
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

For instructions which define objects, the type ID and result ID are
always the second and third SPIR-V instruction word, respectively.

Bug: b/129000021
Change-Id: I6879251732860b80e1f7780d9078ad7bc9751b4c
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43691
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>

This has been removed in LLVM master.

Bug: b/152339534
Change-Id: I613bae6c31944457bd52278228e33af6f85cce97
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43810
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

This patch modifies the CMakeLists.txt file and a few sources
in order to build the SwiftShader Vulkan ICD for Fuchsia with
CMake.

Note that building EGL/GLES_CM/GLESv2 and tests is not supported,
and should be explicitly disabled when configuring the build,
e.g.:

  mkdir build-fuchsia && build-fuchsia
  cmake .. \
    -DCMAKE_TOOLCHAIN_FILE=.../FuchsiaSdkToolchain.cmake \
    -DCMAKE_BUILD_TYPE=Release \
    -DSWIFTSHADER_BUILD_TESTS=0 \
    -DSWIFTSHADER_BUILD_EGL=0 \
    -DSWIFTSHADER_BUILD_GLES_CM=0 \
    -DSWIFTSHADER_BUILD_GLESv2=0 \
    <other-cmake-options>

  make  # or ninja

NOTE: This should not affect the build for other platforms!

This will make it easier to conditionally add Fuchsia-specific
sources to the build (see discussion on b/144687651 for details).

The BUILD.gn file will still be maintained in parallel for
Fuchsia-specific changes.

NOTE: As of now, the Fuchsia SDK doesn't provide a way for
      third-party code to access the framebuffer directly.

      Instead, client code should rely on Vulkan presentation
      surfaces or higher-level APIs provided by the compositor
      that are still in flux / unstable. Hence this will appear
      in a future CL.

Bug: b/143122483
Bug: b/144687651
Change-Id: Ica3b98f5a8eb1370287e2bb4ff22ef938aaa4b49
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/38488Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: David Turner <digit@google.com>

Bug: b/129000021
Change-Id: I0000fc5e65bde87e9037400002db37cb6d50960d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43688
Presubmit-Ready: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>

The targets `libGLESv2` and `libEGL` are no longer public to the root cmake list.
Just move these copy ops into the subdirectory cmake rules without the PVR condition - the copy is cheap enough to do whether we're building PVR or not.

Bug: b/145758253
Change-Id: Ic28331e3797c78e4910332fce934c06c132c3aa8
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43710
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

It correctly points out that despite `Type() = default;`, there's non-assigned const fields, so there's no way to generate a default constructor.

Bug: b/145351270
Change-Id: I1c680982d960df278b5fcca5b29e837a4c82b70d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43709
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Changes:
    de7dffa Socket: Use SO_REUSEADDR, disable SO_LINGER
    9a9d46f Fix bad usage of std::move
    3e6cfd6 CMake: Make the cppdap target's include directory public
    4abe43c optional: Remove the value() method that returns non-const-ref.
    ced82a0 examples: Change stdin and stdout to binary mode
    96b25aa src/io.cpp: Fix uninitialized variable.
    93b8610 examples: Suffix package exe with .exe on Windows
    cdc19ac Serialization: Correctly encode structs with no fields
    3a10d4c Format all source files
    8633aba Add clang-format-all.sh to format all project sources
    1be9bb6 launch.json - fix name of unittests to launch
    d19d8f8 Address #7 nits and fix compiler warning (#9)
    73d697e Fix Response type info, make response 'body' field optional
    eab43f3 net::Server: Fix onError parameter default.

Bug: b/145351270
Change-Id: I6a7767f8b763e8029571d65aa5446f5043ad83ac
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43708
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Following the Vulkan 1.1 specification Section 2.9.1. "Conversion from
Normalized Fixed-Point to Floating-Point":

Note that while zero is exactly expressible in this representation, one
value (-128 in the example) is outside the representable range, and must be
clamped before use.

https://www.khronos.org/registry/vulkan/specs/1.1/html/vkspec.html#fundamentals-fixedfpconv

Bug: dawn:283
Change-Id: Id548c6d67132ac36f33bc020954c5511de624ffc
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43648
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Corentin Wallez <cwallez@google.com>

A test in dEQP was failing using SwANGLE in debug.
The test was using a 1200x1200 grid
 -> 1,44M squares
 -> 2,88M triangles

This was exceeding our hardcoded limit of 1 << 21 (2,097,152).

Removing this code makes the test pass.

Bug: b/152958809
Change-Id: I43fc3c0c9d78378911ab0238a908367ba310c09e
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43668
Presubmit-Ready: Alexis Hétu <sugoi@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Alexis Hétu <sugoi@google.com>

Do this by breaking up the monolithic `runDaily()` function, and combining errors from the new `postDailyResults()` and `postCoverageResults()` functions.

Bug: b/152192800
Change-Id: I031b37fa32d6d05ae1c38dff27a180c809aa4fe1
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43649Tested-by: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

Added CMakeLists.txt to the following folders that produce the
following targets:

src/OpenGL/common -> libGLESCommon (new)
src/OpenGL/compiler -> GLCompiler
src/OpenGL/libEGL -> libEGL
src/OpenGL/libGLESv2 ->libGLESv2
src/OpenGL/libGLES_CM -> libGLES_CM

Change-Id: I8217918b13b6963d18d1e6f89b4fa9e806bb36db
Bug: b/145758253
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43588
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

This change add a CMakeLists in each of the following folders, and
creates the following targets:

src/Common -> gl_common
src/Main -> gl_main
src/Renderer -> gl_renderer
src/Shader -> gl_shader

Bug: b/145758253
Change-Id: Ia1f68d0689f80955586235e125e0197da25d692d
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43550
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Antonio Maiorano <amaiorano@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

This change makes it so that source/Common and source/Main are only
referenced by the SwiftShader target. This will allow for better
target-based sub CMakeLists.

* Remove SharedLibrary.hpp/cpp from source list of libGLESv2 and
  libGLES_CM, as both of these depend on SwiftShader, which already
  builds these files.

* Make libEGL depend on SwiftShader, so we can remove
  SharedLibrary.hpp/cpp and Main/libX11.hpp/cpp from its
  source list.

Bug: b/145758253
Change-Id: If3d8e12cd09fb40d699080c2eeaa8243abe50512
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43528Tested-by: Antonio Maiorano <amaiorano@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>

This benchmarks `sw::LRUCache`, which currently resides in the `src/Device` folder (but will be moved to `src/System`) with 43489.

Bug: b/153338950
Change-Id: I169b6ecb4a35349726a01b1da52472eaef3dfd74
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43492Reviewed-by: Alexis Hétu <sugoi@google.com>
Reviewed-by: Antonio Maiorano <amaiorano@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>

This isn't used.

There's another Cast() declared at vk scope, below VkCommandBuffer that actually is used.

Bug: b/153462569
Change-Id: Id348676c9bb00aa934a972a2833ac6a1d3af54c1
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43575Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Alexis Hétu <sugoi@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

Holds sw::Profiler that is entirely dead code.
There's little here that looks worthy of keeping. Let's remove it.

Bug: b/153462569
Change-Id: I4330781e8ba362461f164f0a52f764d4fe8f4dc4
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43574Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Tested-by: Ben Clayton <bclayton@google.com>

This is a primitive check to ensure that .bp, .gn and .bazel files don't have source references to missing files.

This can catch a common build breakage.

Fixes: b/153439736
Change-Id: I95621f8775a0e536015d4da9897785935f130884
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/43572
Kokoro-Result: kokoro <noreply+kokoro@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Tested-by: Ben Clayton <bclayton@google.com>